The 30% decrease in atmospheric carbon dioxide during glacial maxima must be driven by some change in the chemical circulation of the ocean. Here, a new model for late Quaternary CO2 variability is presented which resolves some problems occurring in previous models (including the timing of carbon dioxide response and changes in the oxygen content of the deep ocean). The primary driving factor in this model is a rearrangement of chemical distributions in the ocean whereby labile nutrients and metabolic CO2 are concentrated in deep waters rather than in intermediate waters as observed in the modern ocean. This new ''bottom-heavy'' chemical structure does not affect atmospheric carbon dioxide directly. Instead, CO2-induced acidity lowers the deep ocean carbonate ion concentration and temporarily increases carbonate dissolution rates. Oceanic alkalinity then rises until the deep ocean carbonate ion is restored to its steady state value. The resulting increase in oceanic alkalinity draws CO2 out of the atmosphere into the ocean. Alkalinity and atmospheric CO2 lag several thousand years behind the change in oceanic chemical structure; this delayed response is determined by the limited rate of continental weathering and deep ocean carbonate dissolution relative to the oceanic alkalinity inventory. It is proposed that characteristic deglacial and preglacial states of the ocean are leading factors driving glacial-interglacial climate changes. This concept reinterprets the carbon isotope contrast between surface water and deep waters <ΔΔ13C(P-B)> as due to a shift of light metabolic carbon from intermediate waters into deep waters. The process is illustrated here by a simple equilibrium five box ocean model. The observed intermediate-depth nutrient depletion is not sufficient in itself to determine which of several possible mechanisms are operating, and some mechanisms are not as effective in changing atmospheric CO2 as others. The largest response is seen from deepening the regeneration cycle of organic carbon; the least response is seen from converting North Atlantic Deep Water into intermediate water. ¿ American Geophysical Union 1988